160630mt

fin2X MT meeting.

Present: Francis, Heiki-Jaan, Tiina, Trond

Agenda:Integrating fin in the fin2X MT systems

Conclusions, in list format:

  • Tag harmonising the fsts:
    • Change fin.fst to sme.fst and est.fst style whenever possible
  • CG outsourcing
    • Do tag conversion (whenever not harmonised) in fst relabeling, and remove from CG.
    • Guesser components in FST and bidix, not in CG
    • ADD (add more) and MAP (add many) - rules for adding syntactic tags is ok
  • Set up Makefiles etc for all directions in apertium
  • Set up transfer rules in .t1x and .lrx
  • CG
    • Find the Finnish objects (ADD @OBJ> and @<OBJ)
  • Tags
    • Whenever tag mismatches (Pron, Adp, ...), omit the finer distinctions
    • Gear the fin tags of indeclinables from morphology to syntax
  • Pronouns
    • Less types, more MT
  • Pcle
    • fin: Pcle, other lgs: Adv and/or Po, Pr
    • The concept of Pcle should be confined to words with more restricted syntax than adverbs

Overall thinking on fin tags:

  • Harmonise form: Pass vs. Pss, Pron vs. Prn to sme / est.
    • Act vs. 0 is not form only: we have Act vs Pass and 0 vs. Der/Pass
  • Change interpretation whenever clear what to do
  • When not, postpone until a , and let MT context be part of decision

Deadlines:

  • Summer: Prepare for first August week
  • First August week: Fix the system (before 1400 Estonian time thursday)